blob: 473e0c0abb8621f732ae3a384acbac63eca8ff11 [file] [log] [blame]
Greg Kroah-Hartmanb2441312017-11-01 15:07:57 +01001/* SPDX-License-Identifier: GPL-2.0 */
Paul Menageddbcc7e2007-10-18 23:39:30 -07002#ifndef _LINUX_CGROUP_H
3#define _LINUX_CGROUP_H
4/*
5 * cgroup interface
6 *
7 * Copyright (C) 2003 BULL SA
8 * Copyright (C) 2004-2006 Silicon Graphics, Inc.
9 *
10 */
11
12#include <linux/sched.h>
Paul Menageddbcc7e2007-10-18 23:39:30 -070013#include <linux/cpumask.h>
14#include <linux/nodemask.h>
Tejun Heoeb6fd502012-11-09 09:12:29 -080015#include <linux/rculist.h>
Balbir Singh846c7bb2007-10-18 23:39:44 -070016#include <linux/cgroupstats.h>
Tejun Heo25a7e682013-04-14 20:15:25 -070017#include <linux/fs.h>
Tejun Heo7da11272013-12-05 12:28:04 -050018#include <linux/seq_file.h>
Tejun Heo2bd59d42014-02-11 11:52:49 -050019#include <linux/kernfs.h>
Tejun Heo49d1dc42015-09-18 11:56:28 -040020#include <linux/jump_label.h>
Aditya Kalia79a9082016-01-29 02:54:06 -060021#include <linux/types.h>
22#include <linux/ns_common.h>
23#include <linux/nsproxy.h>
24#include <linux/user_namespace.h>
Elena Reshetova387ad962017-02-20 12:19:00 +020025#include <linux/refcount.h>
Tejun Heod2cc5ed2017-09-25 08:12:04 -070026#include <linux/kernel_stat.h>
Tejun Heob4a04ab2015-05-13 15:38:40 -040027
28#include <linux/cgroup-defs.h>
Paul Menageddbcc7e2007-10-18 23:39:30 -070029
30#ifdef CONFIG_CGROUPS
31
Tejun Heo6abc8ca2015-08-04 15:20:55 -040032/*
33 * All weight knobs on the default hierarhcy should use the following min,
34 * default and max values. The default value is the logarithmic center of
35 * MIN and MAX and allows 100x to be expressed in both directions.
36 */
37#define CGROUP_WEIGHT_MIN 1
38#define CGROUP_WEIGHT_DFL 100
39#define CGROUP_WEIGHT_MAX 10000
40
Tejun Heobc2fb7e2017-05-15 09:34:01 -040041/* walk only threadgroup leaders */
42#define CSS_TASK_ITER_PROCS (1U << 0)
Tejun Heo450ee0c2017-05-15 09:34:03 -040043/* walk all threaded css_sets in the domain */
44#define CSS_TASK_ITER_THREADED (1U << 1)
Tejun Heobc2fb7e2017-05-15 09:34:01 -040045
Tejun Heoc326aa22015-05-13 16:24:16 -040046/* a css_task_iter should be treated as an opaque object */
47struct css_task_iter {
48 struct cgroup_subsys *ss;
Tejun Heobc2fb7e2017-05-15 09:34:01 -040049 unsigned int flags;
Paul Menageddbcc7e2007-10-18 23:39:30 -070050
Tejun Heoc326aa22015-05-13 16:24:16 -040051 struct list_head *cset_pos;
52 struct list_head *cset_head;
Paul Menageddbcc7e2007-10-18 23:39:30 -070053
Tejun Heo450ee0c2017-05-15 09:34:03 -040054 struct list_head *tcset_pos;
55 struct list_head *tcset_head;
56
Tejun Heoc326aa22015-05-13 16:24:16 -040057 struct list_head *task_pos;
58 struct list_head *tasks_head;
59 struct list_head *mg_tasks_head;
Tejun Heoed27b9f2015-10-15 16:41:52 -040060
61 struct css_set *cur_cset;
Tejun Heo450ee0c2017-05-15 09:34:03 -040062 struct css_set *cur_dcset;
Tejun Heoed27b9f2015-10-15 16:41:52 -040063 struct task_struct *cur_task;
64 struct list_head iters_node; /* css_set->task_iters */
Paul Menage817929e2007-10-18 23:39:36 -070065};
Tejun Heoc326aa22015-05-13 16:24:16 -040066
67extern struct cgroup_root cgrp_dfl_root;
68extern struct css_set init_css_set;
69
70#define SUBSYS(_x) extern struct cgroup_subsys _x ## _cgrp_subsys;
71#include <linux/cgroup_subsys.h>
Paul Menage817929e2007-10-18 23:39:36 -070072#undef SUBSYS
73
Tejun Heo49d1dc42015-09-18 11:56:28 -040074#define SUBSYS(_x) \
75 extern struct static_key_true _x ## _cgrp_subsys_enabled_key; \
76 extern struct static_key_true _x ## _cgrp_subsys_on_dfl_key;
77#include <linux/cgroup_subsys.h>
78#undef SUBSYS
79
80/**
81 * cgroup_subsys_enabled - fast test on whether a subsys is enabled
82 * @ss: subsystem in question
83 */
84#define cgroup_subsys_enabled(ss) \
85 static_branch_likely(&ss ## _enabled_key)
86
87/**
88 * cgroup_subsys_on_dfl - fast test on whether a subsys is on default hierarchy
89 * @ss: subsystem in question
90 */
91#define cgroup_subsys_on_dfl(ss) \
92 static_branch_likely(&ss ## _on_dfl_key)
93
Tejun Heoc326aa22015-05-13 16:24:16 -040094bool css_has_online_children(struct cgroup_subsys_state *css);
95struct cgroup_subsys_state *css_from_id(int id, struct cgroup_subsys *ss);
96struct cgroup_subsys_state *cgroup_get_e_css(struct cgroup *cgroup,
97 struct cgroup_subsys *ss);
98struct cgroup_subsys_state *css_tryget_online_from_dir(struct dentry *dentry,
99 struct cgroup_subsys *ss);
100
Tejun Heo16af4392015-11-20 15:55:52 -0500101struct cgroup *cgroup_get_from_path(const char *path);
Martin KaFai Lau1f3fe7e2016-06-30 10:28:42 -0700102struct cgroup *cgroup_get_from_fd(int fd);
Tejun Heo16af4392015-11-20 15:55:52 -0500103
Tejun Heoc326aa22015-05-13 16:24:16 -0400104int cgroup_attach_task_all(struct task_struct *from, struct task_struct *);
105int cgroup_transfer_tasks(struct cgroup *to, struct cgroup *from);
106
107int cgroup_add_dfl_cftypes(struct cgroup_subsys *ss, struct cftype *cfts);
108int cgroup_add_legacy_cftypes(struct cgroup_subsys *ss, struct cftype *cfts);
109int cgroup_rm_cftypes(struct cftype *cfts);
Tejun Heo34c06252015-11-05 00:12:24 -0500110void cgroup_file_notify(struct cgroup_file *cfile);
Tejun Heoc326aa22015-05-13 16:24:16 -0400111
Tejun Heo4c737b42016-08-10 11:23:44 -0400112int task_cgroup_path(struct task_struct *task, char *buf, size_t buflen);
Tejun Heoc326aa22015-05-13 16:24:16 -0400113int cgroupstats_build(struct cgroupstats *stats, struct dentry *dentry);
114int proc_cgroup_show(struct seq_file *m, struct pid_namespace *ns,
115 struct pid *pid, struct task_struct *tsk);
116
117void cgroup_fork(struct task_struct *p);
Oleg Nesterovb53202e2015-12-03 10:24:08 -0500118extern int cgroup_can_fork(struct task_struct *p);
119extern void cgroup_cancel_fork(struct task_struct *p);
120extern void cgroup_post_fork(struct task_struct *p);
Tejun Heoc326aa22015-05-13 16:24:16 -0400121void cgroup_exit(struct task_struct *p);
Tejun Heo2e91fa72015-10-15 16:41:53 -0400122void cgroup_free(struct task_struct *p);
Tejun Heoc326aa22015-05-13 16:24:16 -0400123
124int cgroup_init_early(void);
125int cgroup_init(void);
126
Tejun Heo5c9d5352014-05-16 13:22:48 -0400127/*
Tejun Heoc326aa22015-05-13 16:24:16 -0400128 * Iteration helpers and macros.
Tejun Heo5c9d5352014-05-16 13:22:48 -0400129 */
Paul Menageddbcc7e2007-10-18 23:39:30 -0700130
Tejun Heoc326aa22015-05-13 16:24:16 -0400131struct cgroup_subsys_state *css_next_child(struct cgroup_subsys_state *pos,
132 struct cgroup_subsys_state *parent);
133struct cgroup_subsys_state *css_next_descendant_pre(struct cgroup_subsys_state *pos,
134 struct cgroup_subsys_state *css);
135struct cgroup_subsys_state *css_rightmost_descendant(struct cgroup_subsys_state *pos);
136struct cgroup_subsys_state *css_next_descendant_post(struct cgroup_subsys_state *pos,
137 struct cgroup_subsys_state *css);
Tejun Heo72c97e52013-08-08 20:11:22 -0400138
Tejun Heo1f7dd3e52015-12-03 10:18:21 -0500139struct task_struct *cgroup_taskset_first(struct cgroup_taskset *tset,
140 struct cgroup_subsys_state **dst_cssp);
141struct task_struct *cgroup_taskset_next(struct cgroup_taskset *tset,
142 struct cgroup_subsys_state **dst_cssp);
Paul Menageddbcc7e2007-10-18 23:39:30 -0700143
Tejun Heobc2fb7e2017-05-15 09:34:01 -0400144void css_task_iter_start(struct cgroup_subsys_state *css, unsigned int flags,
Tejun Heoc326aa22015-05-13 16:24:16 -0400145 struct css_task_iter *it);
146struct task_struct *css_task_iter_next(struct css_task_iter *it);
147void css_task_iter_end(struct css_task_iter *it);
Tejun Heo0ae78e02013-08-13 11:01:54 -0400148
Tejun Heoc326aa22015-05-13 16:24:16 -0400149/**
150 * css_for_each_child - iterate through children of a css
151 * @pos: the css * to use as the loop cursor
152 * @parent: css whose children to walk
153 *
154 * Walk @parent's children. Must be called under rcu_read_lock().
155 *
156 * If a subsystem synchronizes ->css_online() and the start of iteration, a
157 * css which finished ->css_online() is guaranteed to be visible in the
158 * future iterations and will stay visible until the last reference is put.
159 * A css which hasn't finished ->css_online() or already finished
160 * ->css_offline() may show up during traversal. It's each subsystem's
161 * responsibility to synchronize against on/offlining.
162 *
163 * It is allowed to temporarily drop RCU read lock during iteration. The
164 * caller is responsible for ensuring that @pos remains accessible until
165 * the start of the next iteration by, for example, bumping the css refcnt.
166 */
167#define css_for_each_child(pos, parent) \
168 for ((pos) = css_next_child(NULL, (parent)); (pos); \
169 (pos) = css_next_child((pos), (parent)))
Tejun Heod5c419b2014-05-16 13:22:48 -0400170
Tejun Heoc326aa22015-05-13 16:24:16 -0400171/**
172 * css_for_each_descendant_pre - pre-order walk of a css's descendants
173 * @pos: the css * to use as the loop cursor
174 * @root: css whose descendants to walk
175 *
176 * Walk @root's descendants. @root is included in the iteration and the
177 * first node to be visited. Must be called under rcu_read_lock().
178 *
179 * If a subsystem synchronizes ->css_online() and the start of iteration, a
180 * css which finished ->css_online() is guaranteed to be visible in the
181 * future iterations and will stay visible until the last reference is put.
182 * A css which hasn't finished ->css_online() or already finished
183 * ->css_offline() may show up during traversal. It's each subsystem's
184 * responsibility to synchronize against on/offlining.
185 *
186 * For example, the following guarantees that a descendant can't escape
187 * state updates of its ancestors.
188 *
189 * my_online(@css)
190 * {
191 * Lock @css's parent and @css;
192 * Inherit state from the parent;
193 * Unlock both.
194 * }
195 *
196 * my_update_state(@css)
197 * {
198 * css_for_each_descendant_pre(@pos, @css) {
199 * Lock @pos;
200 * if (@pos == @css)
201 * Update @css's state;
202 * else
203 * Verify @pos is alive and inherit state from its parent;
204 * Unlock @pos;
205 * }
206 * }
207 *
208 * As long as the inheriting step, including checking the parent state, is
209 * enclosed inside @pos locking, double-locking the parent isn't necessary
210 * while inheriting. The state update to the parent is guaranteed to be
211 * visible by walking order and, as long as inheriting operations to the
212 * same @pos are atomic to each other, multiple updates racing each other
213 * still result in the correct state. It's guaranateed that at least one
214 * inheritance happens for any css after the latest update to its parent.
215 *
216 * If checking parent's state requires locking the parent, each inheriting
217 * iteration should lock and unlock both @pos->parent and @pos.
218 *
219 * Alternatively, a subsystem may choose to use a single global lock to
220 * synchronize ->css_online() and ->css_offline() against tree-walking
221 * operations.
222 *
223 * It is allowed to temporarily drop RCU read lock during iteration. The
224 * caller is responsible for ensuring that @pos remains accessible until
225 * the start of the next iteration by, for example, bumping the css refcnt.
226 */
227#define css_for_each_descendant_pre(pos, css) \
228 for ((pos) = css_next_descendant_pre(NULL, (css)); (pos); \
229 (pos) = css_next_descendant_pre((pos), (css)))
Tejun Heo15a4c832014-05-04 15:09:14 -0400230
Tejun Heoc326aa22015-05-13 16:24:16 -0400231/**
232 * css_for_each_descendant_post - post-order walk of a css's descendants
233 * @pos: the css * to use as the loop cursor
234 * @css: css whose descendants to walk
235 *
236 * Similar to css_for_each_descendant_pre() but performs post-order
237 * traversal instead. @root is included in the iteration and the last
238 * node to be visited.
239 *
240 * If a subsystem synchronizes ->css_online() and the start of iteration, a
241 * css which finished ->css_online() is guaranteed to be visible in the
242 * future iterations and will stay visible until the last reference is put.
243 * A css which hasn't finished ->css_online() or already finished
244 * ->css_offline() may show up during traversal. It's each subsystem's
245 * responsibility to synchronize against on/offlining.
246 *
247 * Note that the walk visibility guarantee example described in pre-order
248 * walk doesn't apply the same to post-order walks.
249 */
250#define css_for_each_descendant_post(pos, css) \
251 for ((pos) = css_next_descendant_post(NULL, (css)); (pos); \
252 (pos) = css_next_descendant_post((pos), (css)))
Tejun Heo48ddbe12012-04-01 12:09:56 -0700253
Tejun Heoc326aa22015-05-13 16:24:16 -0400254/**
255 * cgroup_taskset_for_each - iterate cgroup_taskset
256 * @task: the loop cursor
Tejun Heo1f7dd3e52015-12-03 10:18:21 -0500257 * @dst_css: the destination css
Tejun Heoc326aa22015-05-13 16:24:16 -0400258 * @tset: taskset to iterate
Tejun Heo4530edd2015-09-11 15:00:19 -0400259 *
260 * @tset may contain multiple tasks and they may belong to multiple
Tejun Heo1f7dd3e52015-12-03 10:18:21 -0500261 * processes.
262 *
263 * On the v2 hierarchy, there may be tasks from multiple processes and they
264 * may not share the source or destination csses.
265 *
266 * On traditional hierarchies, when there are multiple tasks in @tset, if a
267 * task of a process is in @tset, all tasks of the process are in @tset.
268 * Also, all are guaranteed to share the same source and destination csses.
Tejun Heo4530edd2015-09-11 15:00:19 -0400269 *
270 * Iteration is not in any specific order.
Tejun Heoc326aa22015-05-13 16:24:16 -0400271 */
Tejun Heo1f7dd3e52015-12-03 10:18:21 -0500272#define cgroup_taskset_for_each(task, dst_css, tset) \
273 for ((task) = cgroup_taskset_first((tset), &(dst_css)); \
274 (task); \
275 (task) = cgroup_taskset_next((tset), &(dst_css)))
Tejun Heo0cb51d72014-05-16 13:22:49 -0400276
Tejun Heo4530edd2015-09-11 15:00:19 -0400277/**
278 * cgroup_taskset_for_each_leader - iterate group leaders in a cgroup_taskset
279 * @leader: the loop cursor
Tejun Heo1f7dd3e52015-12-03 10:18:21 -0500280 * @dst_css: the destination css
Geliang Tang7b4632f2016-12-24 23:28:35 +0800281 * @tset: taskset to iterate
Tejun Heo4530edd2015-09-11 15:00:19 -0400282 *
283 * Iterate threadgroup leaders of @tset. For single-task migrations, @tset
284 * may not contain any.
285 */
Tejun Heo1f7dd3e52015-12-03 10:18:21 -0500286#define cgroup_taskset_for_each_leader(leader, dst_css, tset) \
287 for ((leader) = cgroup_taskset_first((tset), &(dst_css)); \
288 (leader); \
289 (leader) = cgroup_taskset_next((tset), &(dst_css))) \
Tejun Heo4530edd2015-09-11 15:00:19 -0400290 if ((leader) != (leader)->group_leader) \
291 ; \
292 else
293
Tejun Heoc326aa22015-05-13 16:24:16 -0400294/*
295 * Inline functions.
296 */
Paul Menageddbcc7e2007-10-18 23:39:30 -0700297
Tejun Heo5de01072013-06-12 21:04:52 -0700298/**
299 * css_get - obtain a reference on the specified css
300 * @css: target css
301 *
302 * The caller must already have a reference.
Paul Menageddbcc7e2007-10-18 23:39:30 -0700303 */
Paul Menageddbcc7e2007-10-18 23:39:30 -0700304static inline void css_get(struct cgroup_subsys_state *css)
305{
Tejun Heo3b514d22014-05-16 13:22:47 -0400306 if (!(css->flags & CSS_NO_REF))
307 percpu_ref_get(&css->refcnt);
Paul Menageddbcc7e2007-10-18 23:39:30 -0700308}
Paul Menagee7c5ec92009-01-07 18:08:38 -0800309
Tejun Heo5de01072013-06-12 21:04:52 -0700310/**
Johannes Weinere8ea14c2014-12-10 15:42:42 -0800311 * css_get_many - obtain references on the specified css
312 * @css: target css
313 * @n: number of references to get
314 *
315 * The caller must already have a reference.
316 */
317static inline void css_get_many(struct cgroup_subsys_state *css, unsigned int n)
318{
319 if (!(css->flags & CSS_NO_REF))
320 percpu_ref_get_many(&css->refcnt, n);
321}
322
323/**
Tejun Heo6f4524d2014-05-16 13:22:52 -0400324 * css_tryget - try to obtain a reference on the specified css
325 * @css: target css
326 *
327 * Obtain a reference on @css unless it already has reached zero and is
328 * being released. This function doesn't care whether @css is on or
329 * offline. The caller naturally needs to ensure that @css is accessible
330 * but doesn't have to be holding a reference on it - IOW, RCU protected
331 * access is good enough for this function. Returns %true if a reference
332 * count was successfully obtained; %false otherwise.
333 */
334static inline bool css_tryget(struct cgroup_subsys_state *css)
335{
336 if (!(css->flags & CSS_NO_REF))
337 return percpu_ref_tryget(&css->refcnt);
338 return true;
339}
340
341/**
Tejun Heoec903c02014-05-13 12:11:01 -0400342 * css_tryget_online - try to obtain a reference on the specified css if online
Tejun Heo5de01072013-06-12 21:04:52 -0700343 * @css: target css
344 *
Tejun Heoec903c02014-05-13 12:11:01 -0400345 * Obtain a reference on @css if it's online. The caller naturally needs
346 * to ensure that @css is accessible but doesn't have to be holding a
Tejun Heo5de01072013-06-12 21:04:52 -0700347 * reference on it - IOW, RCU protected access is good enough for this
348 * function. Returns %true if a reference count was successfully obtained;
349 * %false otherwise.
350 */
Tejun Heoec903c02014-05-13 12:11:01 -0400351static inline bool css_tryget_online(struct cgroup_subsys_state *css)
Paul Menagee7c5ec92009-01-07 18:08:38 -0800352{
Tejun Heo3b514d22014-05-16 13:22:47 -0400353 if (!(css->flags & CSS_NO_REF))
354 return percpu_ref_tryget_live(&css->refcnt);
355 return true;
Paul Menagee7c5ec92009-01-07 18:08:38 -0800356}
357
Tejun Heo5de01072013-06-12 21:04:52 -0700358/**
Tejun Heo41c25702017-05-24 12:03:48 -0400359 * css_is_dying - test whether the specified css is dying
360 * @css: target css
361 *
362 * Test whether @css is in the process of offlining or already offline. In
363 * most cases, ->css_online() and ->css_offline() callbacks should be
364 * enough; however, the actual offline operations are RCU delayed and this
365 * test returns %true also when @css is scheduled to be offlined.
366 *
367 * This is useful, for example, when the use case requires synchronous
368 * behavior with respect to cgroup removal. cgroup removal schedules css
369 * offlining but the css can seem alive while the operation is being
370 * delayed. If the delay affects user visible semantics, this test can be
371 * used to resolve the situation.
372 */
373static inline bool css_is_dying(struct cgroup_subsys_state *css)
374{
375 return !(css->flags & CSS_NO_REF) && percpu_ref_is_dying(&css->refcnt);
376}
377
378/**
Tejun Heo5de01072013-06-12 21:04:52 -0700379 * css_put - put a css reference
380 * @css: target css
381 *
Tejun Heoec903c02014-05-13 12:11:01 -0400382 * Put a reference obtained via css_get() and css_tryget_online().
Tejun Heo5de01072013-06-12 21:04:52 -0700383 */
Paul Menageddbcc7e2007-10-18 23:39:30 -0700384static inline void css_put(struct cgroup_subsys_state *css)
385{
Tejun Heo3b514d22014-05-16 13:22:47 -0400386 if (!(css->flags & CSS_NO_REF))
387 percpu_ref_put(&css->refcnt);
Paul Menageddbcc7e2007-10-18 23:39:30 -0700388}
389
Johannes Weinere8ea14c2014-12-10 15:42:42 -0800390/**
391 * css_put_many - put css references
392 * @css: target css
393 * @n: number of references to put
394 *
395 * Put references obtained via css_get() and css_tryget_online().
396 */
397static inline void css_put_many(struct cgroup_subsys_state *css, unsigned int n)
398{
399 if (!(css->flags & CSS_NO_REF))
400 percpu_ref_put_many(&css->refcnt, n);
401}
402
Tejun Heo3e489302017-08-11 05:49:01 -0700403static inline void cgroup_get(struct cgroup *cgrp)
404{
405 css_get(&cgrp->self);
406}
407
408static inline bool cgroup_tryget(struct cgroup *cgrp)
409{
410 return css_tryget(&cgrp->self);
411}
412
Tejun Heo16af4392015-11-20 15:55:52 -0500413static inline void cgroup_put(struct cgroup *cgrp)
414{
415 css_put(&cgrp->self);
416}
417
Tejun Heoc326aa22015-05-13 16:24:16 -0400418/**
419 * task_css_set_check - obtain a task's css_set with extra access conditions
420 * @task: the task to obtain css_set for
421 * @__c: extra condition expression to be passed to rcu_dereference_check()
Paul Menageddbcc7e2007-10-18 23:39:30 -0700422 *
Tejun Heoc326aa22015-05-13 16:24:16 -0400423 * A task's css_set is RCU protected, initialized and exited while holding
424 * task_lock(), and can only be modified while holding both cgroup_mutex
425 * and task_lock() while the task is alive. This macro verifies that the
426 * caller is inside proper critical section and returns @task's css_set.
427 *
428 * The caller can also specify additional allowed conditions via @__c, such
429 * as locks used during the cgroup_subsys::attach() methods.
Paul Menageddbcc7e2007-10-18 23:39:30 -0700430 */
Tejun Heoc326aa22015-05-13 16:24:16 -0400431#ifdef CONFIG_PROVE_RCU
432extern struct mutex cgroup_mutex;
Tejun Heof0d9a5f2015-10-15 16:41:53 -0400433extern spinlock_t css_set_lock;
Tejun Heoc326aa22015-05-13 16:24:16 -0400434#define task_css_set_check(task, __c) \
435 rcu_dereference_check((task)->cgroups, \
436 lockdep_is_held(&cgroup_mutex) || \
Tejun Heof0d9a5f2015-10-15 16:41:53 -0400437 lockdep_is_held(&css_set_lock) || \
Tejun Heoc326aa22015-05-13 16:24:16 -0400438 ((task)->flags & PF_EXITING) || (__c))
439#else
440#define task_css_set_check(task, __c) \
441 rcu_dereference((task)->cgroups)
Tejun Heo2bd59d42014-02-11 11:52:49 -0500442#endif
Paul Menageddbcc7e2007-10-18 23:39:30 -0700443
Tejun Heoc326aa22015-05-13 16:24:16 -0400444/**
445 * task_css_check - obtain css for (task, subsys) w/ extra access conds
446 * @task: the target task
447 * @subsys_id: the target subsystem ID
448 * @__c: extra condition expression to be passed to rcu_dereference_check()
449 *
450 * Return the cgroup_subsys_state for the (@task, @subsys_id) pair. The
451 * synchronization rules are the same as task_css_set_check().
452 */
453#define task_css_check(task, subsys_id, __c) \
454 task_css_set_check((task), (__c))->subsys[(subsys_id)]
455
456/**
457 * task_css_set - obtain a task's css_set
458 * @task: the task to obtain css_set for
459 *
460 * See task_css_set_check().
461 */
462static inline struct css_set *task_css_set(struct task_struct *task)
463{
464 return task_css_set_check(task, false);
465}
466
467/**
468 * task_css - obtain css for (task, subsys)
469 * @task: the target task
470 * @subsys_id: the target subsystem ID
471 *
472 * See task_css_check().
473 */
474static inline struct cgroup_subsys_state *task_css(struct task_struct *task,
475 int subsys_id)
476{
477 return task_css_check(task, subsys_id, false);
478}
479
480/**
Linus Torvaldsbbe179f2015-06-26 19:50:04 -0700481 * task_get_css - find and get the css for (task, subsys)
482 * @task: the target task
483 * @subsys_id: the target subsystem ID
484 *
485 * Find the css for the (@task, @subsys_id) combination, increment a
486 * reference on and return it. This function is guaranteed to return a
487 * valid css.
488 */
489static inline struct cgroup_subsys_state *
490task_get_css(struct task_struct *task, int subsys_id)
491{
492 struct cgroup_subsys_state *css;
493
494 rcu_read_lock();
495 while (true) {
496 css = task_css(task, subsys_id);
497 if (likely(css_tryget_online(css)))
498 break;
499 cpu_relax();
500 }
501 rcu_read_unlock();
502 return css;
503}
504
505/**
Tejun Heoc326aa22015-05-13 16:24:16 -0400506 * task_css_is_root - test whether a task belongs to the root css
507 * @task: the target task
508 * @subsys_id: the target subsystem ID
509 *
510 * Test whether @task belongs to the root css on the specified subsystem.
511 * May be invoked in any context.
512 */
513static inline bool task_css_is_root(struct task_struct *task, int subsys_id)
514{
515 return task_css_check(task, subsys_id, true) ==
516 init_css_set.subsys[subsys_id];
517}
518
519static inline struct cgroup *task_cgroup(struct task_struct *task,
520 int subsys_id)
521{
522 return task_css(task, subsys_id)->cgroup;
523}
Tejun Heoa2dd4242014-03-19 10:23:55 -0400524
Tejun Heo3e489302017-08-11 05:49:01 -0700525static inline struct cgroup *task_dfl_cgroup(struct task_struct *task)
526{
527 return task_css_set(task)->dfl_cgrp;
528}
529
530static inline struct cgroup *cgroup_parent(struct cgroup *cgrp)
531{
532 struct cgroup_subsys_state *parent_css = cgrp->self.parent;
533
534 if (parent_css)
535 return container_of(parent_css, struct cgroup, self);
536 return NULL;
537}
538
Tejun Heob11cfb52015-11-20 15:55:52 -0500539/**
540 * cgroup_is_descendant - test ancestry
541 * @cgrp: the cgroup to be tested
542 * @ancestor: possible ancestor of @cgrp
543 *
544 * Test whether @cgrp is a descendant of @ancestor. It also returns %true
545 * if @cgrp == @ancestor. This function is safe to call as long as @cgrp
546 * and @ancestor are accessible.
547 */
548static inline bool cgroup_is_descendant(struct cgroup *cgrp,
549 struct cgroup *ancestor)
550{
551 if (cgrp->root != ancestor->root || cgrp->level < ancestor->level)
552 return false;
553 return cgrp->ancestor_ids[ancestor->level] == ancestor->id;
554}
555
Sargun Dhillonaed704b2016-08-12 08:56:40 -0700556/**
557 * task_under_cgroup_hierarchy - test task's membership of cgroup ancestry
558 * @task: the task to be tested
559 * @ancestor: possible ancestor of @task's cgroup
560 *
561 * Tests whether @task's default cgroup hierarchy is a descendant of @ancestor.
562 * It follows all the same rules as cgroup_is_descendant, and only applies
563 * to the default hierarchy.
564 */
565static inline bool task_under_cgroup_hierarchy(struct task_struct *task,
566 struct cgroup *ancestor)
567{
568 struct css_set *cset = task_css_set(task);
569
570 return cgroup_is_descendant(cset->dfl_cgrp, ancestor);
571}
572
Tejun Heo07bc3562014-02-13 06:58:39 -0500573/* no synchronization, the result can only be used as a hint */
Tejun Heo27bd4db2015-10-15 16:41:50 -0400574static inline bool cgroup_is_populated(struct cgroup *cgrp)
Tejun Heo07bc3562014-02-13 06:58:39 -0500575{
Tejun Heo454000a2017-05-15 09:34:02 -0400576 return cgrp->nr_populated_csets + cgrp->nr_populated_domain_children +
577 cgrp->nr_populated_threaded_children;
Tejun Heo07bc3562014-02-13 06:58:39 -0500578}
579
Zefan Lif29374b2014-09-19 16:29:31 +0800580/* returns ino associated with a cgroup */
Tejun Heob1664922014-02-11 11:52:49 -0500581static inline ino_t cgroup_ino(struct cgroup *cgrp)
582{
Shaohua Lic53cd492017-07-12 11:49:50 -0700583 return cgrp->kn->id.ino;
Tejun Heob1664922014-02-11 11:52:49 -0500584}
585
Tejun Heob4168642014-05-13 12:16:21 -0400586/* cft/css accessors for cftype->write() operation */
587static inline struct cftype *of_cft(struct kernfs_open_file *of)
Tejun Heo7da11272013-12-05 12:28:04 -0500588{
Tejun Heo2bd59d42014-02-11 11:52:49 -0500589 return of->kn->priv;
Tejun Heo7da11272013-12-05 12:28:04 -0500590}
591
Tejun Heob4168642014-05-13 12:16:21 -0400592struct cgroup_subsys_state *of_css(struct kernfs_open_file *of);
593
594/* cft/css accessors for cftype->seq_*() operations */
595static inline struct cftype *seq_cft(struct seq_file *seq)
596{
597 return of_cft(seq->private);
598}
599
600static inline struct cgroup_subsys_state *seq_css(struct seq_file *seq)
601{
602 return of_css(seq->private);
603}
Tejun Heo59f52962014-02-11 11:52:49 -0500604
Tejun Heoe61734c2014-02-12 09:29:50 -0500605/*
606 * Name / path handling functions. All are thin wrappers around the kernfs
607 * counterparts and can be called under any context.
608 */
609
610static inline int cgroup_name(struct cgroup *cgrp, char *buf, size_t buflen)
611{
Tejun Heofdce6bf2014-03-19 10:23:54 -0400612 return kernfs_name(cgrp->kn, buf, buflen);
Tejun Heoe61734c2014-02-12 09:29:50 -0500613}
614
Tejun Heo4c737b42016-08-10 11:23:44 -0400615static inline int cgroup_path(struct cgroup *cgrp, char *buf, size_t buflen)
Tejun Heoe61734c2014-02-12 09:29:50 -0500616{
Tejun Heofdce6bf2014-03-19 10:23:54 -0400617 return kernfs_path(cgrp->kn, buf, buflen);
Tejun Heoe61734c2014-02-12 09:29:50 -0500618}
619
620static inline void pr_cont_cgroup_name(struct cgroup *cgrp)
621{
Tejun Heofdce6bf2014-03-19 10:23:54 -0400622 pr_cont_kernfs_name(cgrp->kn);
Tejun Heoe61734c2014-02-12 09:29:50 -0500623}
624
625static inline void pr_cont_cgroup_path(struct cgroup *cgrp)
626{
Tejun Heofdce6bf2014-03-19 10:23:54 -0400627 pr_cont_kernfs_path(cgrp->kn);
Tejun Heoe61734c2014-02-12 09:29:50 -0500628}
629
Tejun Heo77f88792017-03-16 16:54:24 -0400630static inline void cgroup_init_kthreadd(void)
631{
632 /*
633 * kthreadd is inherited by all kthreads, keep it in the root so
634 * that the new kthreads are guaranteed to stay in the root until
635 * initialization is finished.
636 */
637 current->no_cgroup_migration = 1;
638}
639
640static inline void cgroup_kthread_ready(void)
641{
642 /*
643 * This kthread finished initialization. The creator should have
644 * set PF_NO_SETAFFINITY if this kthread should stay in the root.
645 */
646 current->no_cgroup_migration = 0;
647}
648
Shaohua Li121508d2017-07-12 11:49:52 -0700649static inline union kernfs_node_id *cgroup_get_kernfs_id(struct cgroup *cgrp)
650{
651 return &cgrp->kn->id;
652}
Shaohua Li69fd5c32017-07-12 11:49:55 -0700653
654void cgroup_path_from_kernfs_id(const union kernfs_node_id *id,
655 char *buf, size_t buflen);
Paul Menageddbcc7e2007-10-18 23:39:30 -0700656#else /* !CONFIG_CGROUPS */
657
Tejun Heof3ba5382015-01-06 12:02:46 -0500658struct cgroup_subsys_state;
Sargun Dhillonaed704b2016-08-12 08:56:40 -0700659struct cgroup;
Tejun Heof3ba5382015-01-06 12:02:46 -0500660
Tejun Heoc326aa22015-05-13 16:24:16 -0400661static inline void css_put(struct cgroup_subsys_state *css) {}
662static inline int cgroup_attach_task_all(struct task_struct *from,
663 struct task_struct *t) { return 0; }
664static inline int cgroupstats_build(struct cgroupstats *stats,
665 struct dentry *dentry) { return -EINVAL; }
666
Paul Menageb4f48b62007-10-18 23:39:33 -0700667static inline void cgroup_fork(struct task_struct *p) {}
Oleg Nesterovb53202e2015-12-03 10:24:08 -0500668static inline int cgroup_can_fork(struct task_struct *p) { return 0; }
669static inline void cgroup_cancel_fork(struct task_struct *p) {}
670static inline void cgroup_post_fork(struct task_struct *p) {}
Li Zefan1ec41832014-03-28 15:22:19 +0800671static inline void cgroup_exit(struct task_struct *p) {}
Tejun Heo2e91fa72015-10-15 16:41:53 -0400672static inline void cgroup_free(struct task_struct *p) {}
Paul Menageddbcc7e2007-10-18 23:39:30 -0700673
Tejun Heoc326aa22015-05-13 16:24:16 -0400674static inline int cgroup_init_early(void) { return 0; }
675static inline int cgroup_init(void) { return 0; }
Tejun Heo77f88792017-03-16 16:54:24 -0400676static inline void cgroup_init_kthreadd(void) {}
677static inline void cgroup_kthread_ready(void) {}
Shaohua Li121508d2017-07-12 11:49:52 -0700678static inline union kernfs_node_id *cgroup_get_kernfs_id(struct cgroup *cgrp)
679{
680 return NULL;
681}
Sridhar Samudralad7926ee2010-05-30 22:24:39 +0200682
Sargun Dhillonaed704b2016-08-12 08:56:40 -0700683static inline bool task_under_cgroup_hierarchy(struct task_struct *task,
684 struct cgroup *ancestor)
685{
686 return true;
687}
Shaohua Li69fd5c32017-07-12 11:49:55 -0700688
689static inline void cgroup_path_from_kernfs_id(const union kernfs_node_id *id,
690 char *buf, size_t buflen) {}
Paul Menageddbcc7e2007-10-18 23:39:30 -0700691#endif /* !CONFIG_CGROUPS */
692
Tejun Heobd1060a2015-12-07 17:38:53 -0500693/*
Tejun Heod2cc5ed2017-09-25 08:12:04 -0700694 * Basic resource stats.
695 */
696#ifdef CONFIG_CGROUPS
697
698#ifdef CONFIG_CGROUP_CPUACCT
699void cpuacct_charge(struct task_struct *tsk, u64 cputime);
700void cpuacct_account_field(struct task_struct *tsk, int index, u64 val);
701#else
702static inline void cpuacct_charge(struct task_struct *tsk, u64 cputime) {}
703static inline void cpuacct_account_field(struct task_struct *tsk, int index,
704 u64 val) {}
705#endif
706
Tejun Heo041cd642017-09-25 08:12:05 -0700707void __cgroup_account_cputime(struct cgroup *cgrp, u64 delta_exec);
708void __cgroup_account_cputime_field(struct cgroup *cgrp,
709 enum cpu_usage_stat index, u64 delta_exec);
710
Tejun Heod2cc5ed2017-09-25 08:12:04 -0700711static inline void cgroup_account_cputime(struct task_struct *task,
712 u64 delta_exec)
713{
Tejun Heo041cd642017-09-25 08:12:05 -0700714 struct cgroup *cgrp;
715
Tejun Heod2cc5ed2017-09-25 08:12:04 -0700716 cpuacct_charge(task, delta_exec);
Tejun Heo041cd642017-09-25 08:12:05 -0700717
718 rcu_read_lock();
719 cgrp = task_dfl_cgroup(task);
720 if (cgroup_parent(cgrp))
721 __cgroup_account_cputime(cgrp, delta_exec);
722 rcu_read_unlock();
Tejun Heod2cc5ed2017-09-25 08:12:04 -0700723}
724
725static inline void cgroup_account_cputime_field(struct task_struct *task,
726 enum cpu_usage_stat index,
727 u64 delta_exec)
728{
Tejun Heo041cd642017-09-25 08:12:05 -0700729 struct cgroup *cgrp;
730
Tejun Heod2cc5ed2017-09-25 08:12:04 -0700731 cpuacct_account_field(task, index, delta_exec);
Tejun Heo041cd642017-09-25 08:12:05 -0700732
733 rcu_read_lock();
734 cgrp = task_dfl_cgroup(task);
735 if (cgroup_parent(cgrp))
736 __cgroup_account_cputime_field(cgrp, index, delta_exec);
737 rcu_read_unlock();
Tejun Heod2cc5ed2017-09-25 08:12:04 -0700738}
739
740#else /* CONFIG_CGROUPS */
741
742static inline void cgroup_account_cputime(struct task_struct *task,
743 u64 delta_exec) {}
744static inline void cgroup_account_cputime_field(struct task_struct *task,
745 enum cpu_usage_stat index,
746 u64 delta_exec) {}
747
748#endif /* CONFIG_CGROUPS */
749
750/*
Tejun Heobd1060a2015-12-07 17:38:53 -0500751 * sock->sk_cgrp_data handling. For more info, see sock_cgroup_data
752 * definition in cgroup-defs.h.
753 */
754#ifdef CONFIG_SOCK_CGROUP_DATA
755
756#if defined(CONFIG_CGROUP_NET_PRIO) || defined(CONFIG_CGROUP_NET_CLASSID)
757extern spinlock_t cgroup_sk_update_lock;
758#endif
759
760void cgroup_sk_alloc_disable(void);
761void cgroup_sk_alloc(struct sock_cgroup_data *skcd);
762void cgroup_sk_free(struct sock_cgroup_data *skcd);
763
764static inline struct cgroup *sock_cgroup_ptr(struct sock_cgroup_data *skcd)
765{
766#if defined(CONFIG_CGROUP_NET_PRIO) || defined(CONFIG_CGROUP_NET_CLASSID)
767 unsigned long v;
768
769 /*
770 * @skcd->val is 64bit but the following is safe on 32bit too as we
771 * just need the lower ulong to be written and read atomically.
772 */
773 v = READ_ONCE(skcd->val);
774
775 if (v & 1)
776 return &cgrp_dfl_root.cgrp;
777
778 return (struct cgroup *)(unsigned long)v ?: &cgrp_dfl_root.cgrp;
779#else
780 return (struct cgroup *)(unsigned long)skcd->val;
781#endif
782}
783
784#else /* CONFIG_CGROUP_DATA */
785
786static inline void cgroup_sk_alloc(struct sock_cgroup_data *skcd) {}
787static inline void cgroup_sk_free(struct sock_cgroup_data *skcd) {}
788
789#endif /* CONFIG_CGROUP_DATA */
790
Aditya Kalia79a9082016-01-29 02:54:06 -0600791struct cgroup_namespace {
Elena Reshetova387ad962017-02-20 12:19:00 +0200792 refcount_t count;
Aditya Kalia79a9082016-01-29 02:54:06 -0600793 struct ns_common ns;
794 struct user_namespace *user_ns;
Eric W. Biedermand08311d2016-08-08 14:25:30 -0500795 struct ucounts *ucounts;
Aditya Kalia79a9082016-01-29 02:54:06 -0600796 struct css_set *root_cset;
797};
798
799extern struct cgroup_namespace init_cgroup_ns;
800
801#ifdef CONFIG_CGROUPS
802
803void free_cgroup_ns(struct cgroup_namespace *ns);
804
805struct cgroup_namespace *copy_cgroup_ns(unsigned long flags,
806 struct user_namespace *user_ns,
807 struct cgroup_namespace *old_ns);
808
Tejun Heo4c737b42016-08-10 11:23:44 -0400809int cgroup_path_ns(struct cgroup *cgrp, char *buf, size_t buflen,
810 struct cgroup_namespace *ns);
Aditya Kalia79a9082016-01-29 02:54:06 -0600811
812#else /* !CONFIG_CGROUPS */
813
814static inline void free_cgroup_ns(struct cgroup_namespace *ns) { }
815static inline struct cgroup_namespace *
816copy_cgroup_ns(unsigned long flags, struct user_namespace *user_ns,
817 struct cgroup_namespace *old_ns)
818{
819 return old_ns;
820}
821
822#endif /* !CONFIG_CGROUPS */
823
824static inline void get_cgroup_ns(struct cgroup_namespace *ns)
825{
826 if (ns)
Elena Reshetova387ad962017-02-20 12:19:00 +0200827 refcount_inc(&ns->count);
Aditya Kalia79a9082016-01-29 02:54:06 -0600828}
829
830static inline void put_cgroup_ns(struct cgroup_namespace *ns)
831{
Elena Reshetova387ad962017-02-20 12:19:00 +0200832 if (ns && refcount_dec_and_test(&ns->count))
Aditya Kalia79a9082016-01-29 02:54:06 -0600833 free_cgroup_ns(ns);
834}
835
Paul Menageddbcc7e2007-10-18 23:39:30 -0700836#endif /* _LINUX_CGROUP_H */